Refactor browser agent and update dependencies

- Updated import statements to use 'patchright' instead of 'playwright'.
- Cleaned up the BrowserUseAgent class for better readability.
- Modified README instructions for browser installation.
- Added new entries to .gitignore for PDF files and workflow.

.gitignore

@@ -187,4 +187,6 @@ data/
 
 # For Config Files (Current Settings)
 .config.pkl
-*.pdf
+*.pdf
+
+workflow

README.md

@@ -68,12 +68,7 @@ uv pip install -r requirements.txt
 Install Browsers in Playwright:
 You can install specific browsers by running:
 ```bash
-playwright install --with-deps chromium
+patchright install chromium
-```
-
-To install all browsers:
-```bash
-playwright install
 ```
 
 #### Step 4: Configure Environment

src/agent/browser_use/browser_use_agent.py

@@ -1,75 +1,37 @@
 from __future__ import annotations
 
 import asyncio
-import gc
-import inspect
-import json
 import logging
 import os
-import re
-import time
-from pathlib import Path
-from typing import Any, Awaitable, Callable, Dict, Generic, List, Optional, TypeVar, Union
-
-from dotenv import load_dotenv
-from langchain_core.language_models.chat_models import BaseChatModel
-from langchain_core.messages import (
-	BaseMessage,
-	HumanMessage,
-	SystemMessage,
-)
 
 # from lmnr.sdk.decorators import observe
-from pydantic import BaseModel, ValidationError
-
 from browser_use.agent.gif import create_history_gif
-from browser_use.agent.memory.service import Memory, MemorySettings
-from browser_use.agent.message_manager.service import MessageManager, MessageManagerSettings
-from browser_use.agent.message_manager.utils import convert_input_messages, extract_json_from_model_output, save_conversation
-from browser_use.agent.prompts import AgentMessagePrompt, PlannerPrompt, SystemPrompt
-from browser_use.agent.views import (
-	REQUIRED_LLM_API_ENV_VARS,
-	ActionResult,
-	AgentError,
-	AgentHistory,
-	AgentHistoryList,
-	AgentOutput,
-	AgentSettings,
-	AgentState,
-	AgentStepInfo,
-	StepMetadata,
-	ToolCallingMethod,
-)
-from browser_use.browser.browser import Browser
-from browser_use.browser.context import BrowserContext
-from browser_use.browser.views import BrowserState, BrowserStateHistory
-from browser_use.controller.registry.views import ActionModel
-from browser_use.controller.service import Controller
-from browser_use.dom.history_tree_processor.service import (
-	DOMHistoryElement,
-	HistoryTreeProcessor,
-)
-from browser_use.exceptions import LLMException
-from browser_use.telemetry.service import ProductTelemetry
-from browser_use.telemetry.views import (
-	AgentEndTelemetryEvent,
-	AgentRunTelemetryEvent,
-	AgentStepTelemetryEvent,
-)
-from browser_use.utils import check_env_variables, time_execution_async, time_execution_sync
 from browser_use.agent.service import Agent, AgentHookFunc
+from browser_use.agent.views import (
+    AgentHistoryList,
+    AgentStepInfo,
+)
+from browser_use.telemetry.views import (
+    AgentEndTelemetryEvent,
+)
+from browser_use.utils import time_execution_async
+from dotenv import load_dotenv
 
 load_dotenv()
 logger = logging.getLogger(__name__)
 
-SKIP_LLM_API_KEY_VERIFICATION = os.environ.get('SKIP_LLM_API_KEY_VERIFICATION', 'false').lower()[0] in 'ty1'
+SKIP_LLM_API_KEY_VERIFICATION = (
+    os.environ.get("SKIP_LLM_API_KEY_VERIFICATION", "false").lower()[0] in "ty1"
+)
 
 
 class BrowserUseAgent(Agent):
-    @time_execution_async('--run (agent)')
+    @time_execution_async("--run (agent)")
     async def run(
-            self, max_steps: int = 100, on_step_start: AgentHookFunc | None = None,
+        self,
-            on_step_end: AgentHookFunc | None = None
+        max_steps: int = 100,
+        on_step_start: AgentHookFunc | None = None,
+        on_step_end: AgentHookFunc | None = None,
     ) -> AgentHistoryList:
         """Execute the task with maximum number of steps"""
 
@@ -88,7 +50,7 @@ class BrowserUseAgent(Agent):
         signal_handler.register()
 
         # Wait for verification task to complete if it exists
-        if hasattr(self, '_verification_task') and not self._verification_task.done():
+        if hasattr(self, "_verification_task") and not self._verification_task.done():
             try:
                 await self._verification_task
             except Exception:
@@ -100,7 +62,9 @@ class BrowserUseAgent(Agent):
 
             # Execute initial actions if provided
             if self.initial_actions:
-                result = await self.multi_act(self.initial_actions, check_for_new_elements=False)
+                result = await self.multi_act(
+                    self.initial_actions, check_for_new_elements=False
+                )
                 self.state.last_result = result
 
             for step in range(max_steps):
@@ -112,12 +76,14 @@ class BrowserUseAgent(Agent):
 
                 # Check if we should stop due to too many failures
                 if self.state.consecutive_failures >= self.settings.max_failures:
-                    logger.error(f'❌ Stopping due to {self.settings.max_failures} consecutive failures')
+                    logger.error(
+                        f"❌ Stopping due to {self.settings.max_failures} consecutive failures"
+                    )
                     break
 
                 # Check control flags before each step
                 if self.state.stopped:
-                    logger.info('Agent stopped')
+                    logger.info("Agent stopped")
                     break
 
                 while self.state.paused:
@@ -142,13 +108,15 @@ class BrowserUseAgent(Agent):
                     await self.log_completion()
                     break
             else:
-                logger.info('❌ Failed to complete task in maximum steps')
+                logger.info("❌ Failed to complete task in maximum steps")
 
             return self.state.history
 
         except KeyboardInterrupt:
             # Already handled by our signal handler, but catch any direct KeyboardInterrupt as well
-            logger.info('Got KeyboardInterrupt during execution, returning current history')
+            logger.info(
+                "Got KeyboardInterrupt during execution, returning current history"
+            )
             return self.state.history
 
         finally:
@@ -171,8 +139,10 @@ class BrowserUseAgent(Agent):
             await self.close()
 
             if self.settings.generate_gif:
-                output_path: str = 'agent_history.gif'
+                output_path: str = "agent_history.gif"
                 if isinstance(self.settings.generate_gif, str):
                     output_path = self.settings.generate_gif
 
-                create_history_gif(task=self.task, history=self.state.history, output_path=output_path)
+                create_history_gif(
+                    task=self.task, history=self.state.history, output_path=output_path
+                )

src/browser/custom_browser.py

@@ -1,17 +1,17 @@
 import asyncio
 import pdb
 
-from playwright.async_api import Browser as PlaywrightBrowser
+from patchright.async_api import Browser as PlaywrightBrowser
-from playwright.async_api import (
+from patchright.async_api import (
     BrowserContext as PlaywrightBrowserContext,
 )
-from playwright.async_api import (
+from patchright.async_api import (
     Playwright,
     async_playwright,
 )
 from browser_use.browser.browser import Browser, IN_DOCKER
 from browser_use.browser.context import BrowserContext, BrowserContextConfig
-from playwright.async_api import BrowserContext as PlaywrightBrowserContext
+from patchright.async_api import BrowserContext as PlaywrightBrowserContext
 import logging
 
 from browser_use.browser.chrome import (

src/browser/custom_context.py

@@ -4,8 +4,8 @@ import os
 
 from browser_use.browser.browser import Browser, IN_DOCKER
 from browser_use.browser.context import BrowserContext, BrowserContextConfig
-from playwright.async_api import Browser as PlaywrightBrowser
+from patchright.async_api import Browser as PlaywrightBrowser
-from playwright.async_api import BrowserContext as PlaywrightBrowserContext
+from patchright.async_api import BrowserContext as PlaywrightBrowserContext
 from typing import Optional
 from browser_use.browser.context import BrowserContextState
 

tests/test_agents.py

@@ -169,7 +169,7 @@ async def test_browser_use_agent():
 async def test_browser_use_parallel():
     from browser_use.browser.context import BrowserContextWindowSize
     from browser_use.browser.browser import BrowserConfig
-    from playwright.async_api import async_playwright
+    from patchright.async_api import async_playwright
     from browser_use.browser.browser import Browser
     from src.browser.custom_context import BrowserContextConfig
     from src.controller.custom_controller import CustomController

tests/test_playwright.py

@@ -6,7 +6,7 @@ load_dotenv()
 
 def test_connect_browser():
     import os
-    from playwright.sync_api import sync_playwright
+    from patchright.sync_api import sync_playwright
 
     chrome_exe = os.getenv("CHROME_PATH", "")
     chrome_use_data = os.getenv("CHROME_USER_DATA", "")